Counterfactual legislatures

What if the districts were drawn by an algorithm that knows nothing about politics?

US congressional districts are drawn by state legislatures — the people who then run in them. The shortest splitline method removes every human decision: cut the state in two with the shortest straight line that leaves the right number of people on each side, then recurse. No communities of interest, no incumbents, no voting history. Just population and geometry.

These maps are rebuilt from the 2000, 2010 and 2020 censuses — about 8 million census blocks per run. Brightness is population density, so the cities show up as the bright speckle.

The contiguous United States, 432 districts

Shortest-splitline districting of the contiguous United States, 432 districts, 2020 census
2020 census. State lines ignored entirely — the country is cut as one region. Alaska and Hawaii keep their own maps, so 435 − 3 = 432 districts here. The first cut runs from Michigan to Louisiana.

One cut at a time

Every map is a binary tree of straight lines, so the whole thing can be replayed:

Animation of Colorado being divided into 8 districts
Colorado, 8 districts. The Denver metro takes half of them.
Animation of Maryland being divided into 8 districts
Maryland, 8 districts — currently home to some of the most contorted districts in the country.

Does it reproduce the original?

The Center for Range Voting published splitline maps from the 2000 census in 2009. Rebuilding them from scratch is the test of whether any of this is right — same census, same seat counts, independent code:

Center for Range Voting's 2009 splitline map of Colorado
Center for Range Voting, 2009.
This project's splitline map of Colorado, 2000 census
This project, rebuilt from the raw 2000 census blocks. 97.2% of the state's land area falls in the same district in both.

Colorado is the cleanest comparison because it is landlocked and nearly rectangular: there is no water for two implementations to disagree about. Elsewhere they can, because the length of a cut can be measured two ways — the straight-line span, or only the land it crosses. Nothing decides more. Measured as span, the very first national cut is a tilted line from Michigan's Lower Peninsula down to the Louisiana coast west of New Orleans, passing eight kilometres from downtown Indianapolis. Measured as land, it swings to a near-vertical line from the Upper Peninsula across Lake Michigan to the Florida–Alabama border. Since the algorithm always takes the shortest cut, that one definition decides which cut wins — and a single different cut at the root rearranges every one of the 432 districts beneath it.

A few states

Texas divided into 38 districts
Texas — mid-decade redrawn in 2025, and in court over it since.
California divided into 52 districts
California — half its districts sit in the Los Angeles and Bay Area knots.

What this is for

Not a proposal. Splitline districts ignore everything people reasonably care about, and multi-member proportional representation is almost certainly a better answer. The point is that a politically blind map is a measuring stick: the distance between it and the enacted map is a floor on how much the real map was shaped by something other than geography.

That question got sharper in April 2026, when the Supreme Court held in Louisiana v. Callais that a state may draw districts for any political purpose regardless of racial effect. The defence is now "our intent was partisan, not racial" — and a map drawn with no political input at all is the natural thing to measure it against.

The longer aim is to do similar exercises for several countries and electoral systems, and then ask whether legislatures elected under different rules would have voted differently on policies whose effects we can actually measure.